Blood transcriptome marker, kit, system, method and application for FDG-PET metabolic load evaluation

Abstract

The invention discloses a blood transcriptome marker, a kit, a system, a method and application for FDG-PET metabolic load evaluation. The invention provides a blood substitution index of focus FDG-PET inflammation intensity, the substitution index comprises blood transcriptome markers GAS6 and GBP6 and a linear scoring model constructed based on the blood transcriptome markers GAS6 and GBP6, and the substitution index can be used for estimating or reflecting the metabolic load of a focus level in a layering manner and can be used for clinical diagnosis shunt and curative effect follow-up visit. According to the method, a small number of transcripts with the highest positive correlation degree with the focus SUVmax are identified in training data, a linear scoring model is constructed, the focus SUVmax / TGA is mapped and estimated according to the score, and / or whether an active focus exists or not is judged according to a threshold value. The scoring model is strongly and positively correlated with PET image indexes, and can be used as an alternative index of PET metabolic load to assist clinical lesion activity assessment and diagnosis shunting.

Description

Technical Field

[0001] This invention relates to the fields of molecular diagnostics and bioinformatics, specifically to blood transcriptome biomarkers, kits, systems, methods, and applications for FDG-PET metabolic burden assessment, for early screening of infectious diseases, development of personalized intervention strategies, and tuberculosis risk stratification and preventive treatment decisions in public health surveillance systems. Background Technology

[0002] Currently, PET-CT-based imaging is a commonly used method for the preliminary clinical diagnosis of pulmonary tuberculosis. The roles and functions of PET indicators are as follows: 18F-FDG PET-CT can assess the metabolic activity of tuberculosis lesions. TGA (total glycolytic activity) comprehensively reflects the overall metabolic burden of lesions and is more suitable for monitoring treatment response / efficacy; SUVmax emphasizes the peak inflammatory intensity of the most active lesions and has unique value in stratifying baseline lesion activity and measuring peak inflammation. The two are complementary in their characterization dimensions.

[0003] Reported blood transcriptome signatures (such as RISK6 and Sweeney3) mostly correspond to the overall burden dimension of TGA, which is more suitable for applications such as efficacy tracking; however, there is currently a lack of peripheral blood molecule readouts that are dedicated to SUVmax, have a minimal structure, and can be standardized, so as to achieve equivalent stratification and measurement of peak inflammatory lesions without imaging conditions, especially in primary care / resource-constrained and batch screening scenarios.

[0004] QuantiFERON (QFT) is an interferon-gamma release assay (IGRA). Peripheral blood is stimulated with specific antigens such as ESAT-6 / CFP-10, and the quantitative reading (IU / mL) of induced IFN-γ is measured to identify tuberculosis-specific immune responses (evidence of infection). QFT is not a direct indicator of lesion activity, but its quantitative level reflects the degree of systemic immune activation. Literature shows a significant positive correlation between the quantitative QFT at initial diagnosis and the SUVmax of initial PET / CT, suggesting a correspondence between SUVmax and the body's specific T-cell response, and that it can serve as a molecular target for peak inflammatory intensity.

[0005] However, there is currently a lack of peripheral blood molecular markers with a minimal structure that can stably represent SUVmax in imaging-free conditions. It is difficult to achieve equivalent measurement and stratification of peak inflammatory lesions at baseline. Furthermore, there are still cost and complexity obstacles in standardization and qPCR / POC deployment, which have become key bottlenecks for the implementation of imaging-molecular integration in imaging-free / large-scale scenarios. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a two-gene, easily standardized blood transcriptome biomarker (PET2) with SUVmax as the specific target. It enables direct replacement measurement of peak inflammation in lesions and baseline stratification / disease assessment in the absence of PET / CT. At the same time, it maintains complementarity with existing panels based on TGA in application scenarios (treatment monitoring vs. peak inflammation measurement) and significantly reduces implementation complexity, facilitating low-cost, large-scale deployment in qPCR / POC format.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a blood transcriptome biomarker for assessing the metabolic burden of tuberculosis by FDG-PET, wherein the blood transcriptome biomarker is a combination of GAS6 and GBP6.

[0008] In a second aspect, the present invention provides the use of blood transcriptome biomarkers or their detection reagents as described in the first aspect, the use being selected from: i) for the preparation of products for assessing or stratifying the metabolic burden at the level of tuberculosis lesions; ii) for the preparation of products for the auxiliary diagnosis of active pulmonary tuberculosis (PTB) and extrapulmonary tuberculosis (EPTB); iii) for the preparation of products for assessing the metabolic inflammation intensity of pulmonary tuberculosis lesions; iv) for the preparation of products for assessing the activity of pulmonary tuberculosis lesions; v) for the preparation of products for predicting the risk of tuberculosis progression; and vi) for the preparation of products for monitoring the response or efficacy of pulmonary tuberculosis treatment.

[0009] In some implementations, the product includes a reagent kit, a detection chip, a system, or a device.

[0010] In some embodiments, the detection reagent includes reagents for blood transcriptome markers at the gene and / or protein levels.

[0011] In some embodiments, the detection reagent is a reagent selected from one or more detection techniques or methods chosen from the group consisting of: enzyme-linked immunosorbent assay (ELISA), immunofluorescence assay, radioimmunoassay, immunoprecipitation assay, Western blotting, high performance liquid chromatography (HPLC), capillary gel electrophoresis, near-infrared spectroscopy, mass spectrometry, immunochemiluminescence assay, colloidal gold immunochromatography, fluorescence immunochromatography, surface plasmon resonance (SPR), immuno-PCR, and biotin-avidin assay.

[0012] In some implementations, the subjects targeted in v) are 18-20 years old.

[0013] Thirdly, the present invention provides a kit for assessing the metabolic burden of tuberculosis FDG-PET, the kit comprising at least reagents for detecting the expression levels of genes or proteins of blood transcriptome markers as described in the first aspect.

[0014] In some embodiments, the kit includes at least a detection reagent for specifically detecting the mRNA expression level of the blood transcriptome marker.

[0015] In some embodiments, the kit is evaluated using the following method: a scoring model PET2 = z(GAS6) + z(GBP6), where z represents the z-score normalization of the expression level of the corresponding gene.

[0016] Fourthly, the present invention provides a system for assessing the metabolic burden of tuberculosis patients using FDG-PET, the system comprising: The data input module is used to receive the expression level data of the subject's blood transcriptome markers, wherein the blood transcriptome markers are the blood transcriptome markers as described in claim 1; The data processing module stores trained machine learning classification models, which are used to process data and output classification results; The results output module is used to output the FDG-PET metabolic load assessment results of the subjects.

[0017] Fifthly, the present invention provides a method for assessing the metabolic burden of tuberculosis patients using FDG-PET, the method being implemented using the system described in the fourth aspect, comprising the following steps: a) Obtain blood samples from the subjects; b) Detect the expression levels of the blood transcriptome markers in blood samples; c) Input the expression level data obtained in b) into a trained machine learning scoring model, which outputs the subject's FDG-PET metabolic load assessment results.

[0018] In some implementations, the machine learning scoring model is a linear scoring model constructed based on the gene expression levels of the blood transcriptome markers: PET2 = z(GAS6) + z(GBP6), where z represents the z-score normalization of the expression level of the corresponding gene.

[0019] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention provides a linear scoring model (PET2) calculated based on a small number of blood-expressed genes (two genes are preferred embodiments) to reflect the metabolic inflammation intensity of pulmonary tuberculosis lesions; the PET2 can be used for the auxiliary diagnosis of active pulmonary tuberculosis (PTB) and extrapulmonary tuberculosis (EPTB), and has good ability to distinguish from healthy controls, with AUC exceeding 95%.

[0020] (2) The scoring model is strongly positively correlated with PET imaging indicators (especially SUVmax and TGA), and can be used as a substitute indicator for PET metabolic load to assist clinical assessment of lesion activity and diagnostic triage.

[0021] (3) Although PET2 has an age-dependent limitation in predicting the risk of progression, it performs well in PET coupling and distinguishing non-tuberculous diseases, and is suitable as a biological characterization of lesion activity. Attached Figure Description

[0022] Figure 1 PET inflammation coupling analysis: Scatter plots showing the correlation between PET2, RISK6, and Sweeney3 scores and lesion SUVmax (top row) and TGA (bottom row) before and at the end of treatment. The lines show linear fit and confidence intervals. PET2 showed the strongest coupling relationship with PET indicators, and RISK6 and Sweeney3 also showed positive correlation.

[0023] Figure 2 Diagnostic behavior in tuberculosis and non-tuberculosis diseases: ROC curves (left figure for each group) and violin plots (right figure) demonstrate the sensitivity of PET2, RISK6, and Sweeney3 in distinguishing tuberculosis (PTB, EPTB) from healthy controls, as well as their specificity differences compared to non-tuberculosis diseases (granulomatous disease, influenza, lupus erythematosus, rheumatoid arthritis). PET2 and RISK6 showed similar performance in granulomatous disease and influenza, but were more specific in lupus erythematosus and rheumatoid arthritis; Sweeney3 showed the best specificity in influenza.

[0024] Figure 3 The predictive ability of progression in age stratification was assessed. Based on the ACS, Leicester, and London cohorts, ROC curves for PET2, RISK6, and Sweeney3 were plotted for the preset age groups within a 0–12 month prediction window, with guide lines and confidence intervals added. Results showed that PET2 only matched RISK6 in the 18–20 year age group, with performance declining in other age groups. Dashed boxes represent the WHO TPP minimum standard (75% sensitivity / 75% specificity), and diagonal dashed lines represent no discriminative power. Detailed Implementation

[0025] To make the present invention more apparent and understandable, preferred embodiments are described in detail below with reference to the accompanying drawings.

[0026] Example 1: Screening and scoring model construction of molecular biomarkers for PET index substitution This invention provides a "blood replacement index for lesion FDG-PET inflammation intensity" to estimate or stratify the metabolic load at the lesion level (represented by SUVmax and / or total glycolytic activity TGA), and can be used for clinical diagnostic triage and efficacy follow-up. The core approach is to identify a small number of transcripts with the highest positive correlation to lesion SUVmax in the training data, construct a simplified linear scoring model (two genes preferred), use this score to estimate lesion SUVmax / TGA, and / or distinguish between active and inactive lesions based on thresholds. The screening and scoring model construction process is as follows: 1. Training data and feature metrics: Expression profiles and TGA information provided in the GSE89403 dataset, along with SUVmax data from the same subject cohort (Source: Nat Med. 2016 Oct;22(10):1094-1100. doi: 10.1038 / nm.4177. Supplementary Dataset 6), were used for modeling analysis. For each subject, the SUVmax and TGA values ​​of the most significant inflammatory lesions were obtained and matched with their expression profiles (from peripheral blood whole blood RNA expression data). Using SUVmax as the primary training target indicator, the correlation coefficients between whole transcript expression and SUVmax were calculated (Spearman ρ preferred), and the significance of the correlation (e.g., p-value after FDR correction) was recorded as the basis for feature selection and model input. For subsets with TGA data, the same analysis could be performed in parallel to provide alternative target modeling references.

[0027] 2. PET Mapping and Calibration: To estimate the metabolic burden of lesions (such as SUVmax or TGA) from PET2 mapping, the following strategy was adopted: First, discovery and validation subsets were constructed from samples whose expression profiles matched PET data. In the discovery subset, the Spearman correlation coefficient and p-value between the expression levels of all genes at the diagnostic time point and SUVmax were calculated, recorded, and sorted by absolute correlation strength to select the top 30 candidate genes.

[0028] Subsequently, a repeatability correlation test was performed on these 30 candidate genes in the validation subset, and genes with high consistency that were significantly correlated in both subsets were selected. Based on these top genes, all possible pairwise combinations were constructed, and the correlation coefficient between the mean expression of each pair in the entire sample and SUVmax was calculated. Finally, the gene pair with the highest correlation to SUVmax was selected, and its normalized expression sum was defined as the molecular score PET2.

[0029] 3. Candidate gene screening and two-gene score construction: Genes with the highest to lowest positive correlation with SUVmax were selected, and their expression was standardized (e.g., between-subject z-score). The score "PET2" for both genes was defined as the sum of their standardized expression: PET2 = z(GAS6) + z(GBP6). This score must simultaneously satisfy the following conditions in the training set: it must be significantly positively correlated with SUVmax / TGA.

[0030] 4. Diagnostic Thresholds and Usage Rules: The ROC / Youden index is used to determine the diagnostic thresholds for PET2 on the training set for triage between "suspected / confirmed tuberculosis vs. healthy controls (or low-inflammatory states)". For the "lesion activity" stratification, the clinical reference interval and risk stratification threshold for the estimated SUVmax value can be set based on the PET mapping results (e.g., estimated SUVmax ≥ S). ＊ Classified as "highly active"). Its triage efficacy and robustness were verified in independent data, reporting sensitivity, specificity, AUC, and 95% CI.

[0031] Outputs and software modules: This invention provides supporting software modules to complete the following steps: data import → expression standardization → PET2 calculation → PET mapping estimation and confidence interval → diagnostic threshold determination / risk stratification → follow-up trend chart and report export. The modules include parameter encryption and calibration update interfaces for easy local recalibration and version management.

[0032] Alternative implementation methods: Reasonable changes to the following elements are permitted without compromising the implementation effect: (i) number of genes (1 or ≥3); (ii) target image indicators (TGA, SUVpeak, MTV); (iii) correlation measurement method (Kendall τ, etc.); (iv) modeling method (sparse regression, multivariate principal component analysis, etc.).

[0033] Supplementary information on data sources (for citation in other materials / background technology): GSE89403 provides expression profiles and associated TGA indices from lavage samples of patients with active tuberculosis; Nat Med. 2016 Oct;22(10):1094-1100 provides SUVmax data for the same subject cohort (see its supplementary dataset 6). This invention constructs molecular alternative models for PET indices based on these two publicly available data sources, without using blood samples.

[0034] Example 2: Coupled Analysis of Inflammatory Markers in PET In the GSE89403 dataset, based on transcriptomic data at the diagnosis time point (before treatment), we assessed the coupling strength between PET2 and lesion metabolic burden indicators (SUVmax and TGA), and compared it with two published tuberculosis transcriptomic scoring models, RISK6 and Sweeney3. The analysis employed Spearman correlation assessment between whole blood expression data (log2 transformed, standardized) and imaging indicators (SUVmax, TGA), and was presented using linear regression plots.

[0035] The results are as follows Figure 1 As shown, Figure 1 The results showed that the Spearman correlation coefficient between PET2 and SUVmax was R = 0.527 (p = 9.28e-08), and the correlation with TGA was even stronger, at R = 0.619 (p ≈ 0), indicating the strongest coupling relationship among the three scoring models. In contrast, RISK6 and Sweeney3 also showed positive correlations with the PET index, but the strength was weaker (R < 0.52).

[0036] These results indicate that PET2 better reflects the intensity of metabolic inflammation at the lesion level and has the potential to serve as a PET indicator of blood replacement volume.

[0037] Example 3: Comparison of diagnostic performance and specificity for non-tuberculous diseases This embodiment systematically compares the diagnostic performance and specificity of three scoring models—PET2, RISK6, and Sweeney3—in active pulmonary tuberculosis (PTB), extrapulmonary tuberculosis (EPTB), and various non-tuberculous diseases (granulomatous disease, rheumatoid arthritis, systemic lupus erythematosus, and influenza). The results are as follows: Figure 2 As shown.

[0038] Figure 2 The results showed that in the "tuberculosis vs. healthy controls" study, the AUC of PET2 was 96.3% (95% CI: 92.2–100.0), which was similar to that of RISK6 (98.6%) and Sweeney3 (97.8%); in the "tuberculosis vs. healthy controls" study, PET2 also reached 95.8%.

[0039] Figure 2The results showed that for non-tuberculous diseases, PET2 exhibited high cross-reactivity in granulomatous disease and influenza (98.1% and 85.4%, respectively), but higher specificity in systemic lupus erythematosus (74.3%) and rheumatoid arthritis (57.9%), which was superior to RISK6 (83.2% and 68.6%, respectively). Sweeney3 had the highest specificity in influenza (36.6%), but its cross-reactivity in other diseases was relatively high.

[0040] These results suggest that PET2 and RISK6 exhibit similar specificity in most scenarios, but PET2 demonstrates superior specificity in lupus and rheumatoid arthritis, providing direction for the future development of non-cross-reactive screening tools.

[0041] The datasets used include: PTB, EPTB, and influenza patient data from GSE111368; granulomatous disease data from a subset of GSE111368; systemic lupus erythematosus (SLE) data from GSE72509; and rheumatoid arthritis (RA) data from GSE117769.

[0042] Example 4: Performance evaluation of age-stratified tuberculosis progression prediction This embodiment compared the predictive performance of PET2, RISK6, and Sweeney3 in tuberculosis progression across different age groups using data from the ACS, Leicester, and London cohorts within a 0-12 month prediction window. Participants were divided into four age groups: 12-17 years, 18-20 years, 21-35 years, and 36-84 years. Results are as follows: Figure 3 As shown.

[0043] Figure 3 The results showed that PET2 had the strongest predictive power in the 18-20 age group, with an AUC of 97.8% (95% CI: 93.7-100.0), which was roughly on par with RISK6 and Sweeney3. In the other three age groups, the AUC of PET2 decreased significantly. 12-17 years old: AUC = 73.8%; 21-35 years old: AUC = 54.6%; 36-84 years old: AUC = 48.1%.

[0044] In comparison, RISK6 performs relatively well across all age groups, while Sweeney3 falls somewhere in between.

[0045] The results above indicate that although PET2 is strongly coupled with PET activity indicators, its performance in predicting "future progression risk" is significantly age-dependent, only comparable to the current standard model in the 18-20 age group, and inadequate in other age groups. Therefore, this model is more suitable as a tool for characterizing the inflammatory state of lesions than as a general progression prediction model.

[0046] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any form or substance. It should be noted that those skilled in the art can make several improvements and additions without departing from the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. A blood transcriptome biomarker for assessing the metabolic burden of tuberculosis patients using FDG-PET, characterized in that, The blood transcriptome markers are a combination of GAS6 and GBP6.

2. The use of the blood transcriptome biomarker or its detection reagent according to claim 1, characterized in that, The intended uses are selected from: i) products for assessing or stratifying the metabolic burden at the level of tuberculosis lesions; ii) products for aiding in the diagnosis of active pulmonary tuberculosis (PTB) and extrapulmonary tuberculosis (EPTB); iii) products for assessing the intensity of metabolic inflammation in pulmonary tuberculosis lesions; iv) products for assessing the activity of pulmonary tuberculosis lesions; v) products for predicting the risk of tuberculosis progression; and vi) products for monitoring the response to or efficacy of tuberculosis treatment.

3. The application as described in claim 2, characterized in that, The products include reagent kits, detection chips, systems, or devices.

4. The application as described in claim 2, characterized in that, The detection reagent contains reagents for blood transcriptome markers at the gene and / or protein levels.

5. The application as described in claim 4, characterized in that, The detection reagent is a reagent selected from one or more detection techniques or methods from the group consisting of: enzyme-linked immunosorbent assay (ELISA), immunofluorescence assay, radioimmunoassay, immunoprecipitation assay, immunoblotting, high performance liquid chromatography (HPLC), capillary gel electrophoresis, near-infrared spectroscopy, mass spectrometry, immunochemiluminescence assay, colloidal gold immunochromatography, fluorescence immunochromatography, surface plasmon resonance (SPR), immuno-PCR, and biotin-avidin assay.

6. A kit for assessing the metabolic burden of tuberculosis patients using FDG-PET, characterized in that, The kit includes at least one reagent for detecting the expression levels of the blood transcriptome markers described in claim 1.

7. The kit according to claim 6, characterized in that, The kit includes at least one detection reagent for specifically detecting the mRNA expression level of the blood transcriptome marker.

8. The reagent kit as described in claim 6, characterized in that, The kit assesses FDG-PET metabolic load using the scoring model PET2 = z(GAS6) + z(GBP6), where z represents the z-score normalization of the expression level of the corresponding gene.

9. A system for assessing the metabolic burden of tuberculosis patients using FDG-PET, characterized in that, include: The data input module is used to receive the expression level data of the subject's blood transcriptome markers, wherein the blood transcriptome markers are the blood transcriptome markers as described in claim 1; The data processing module stores trained machine learning scoring models, which are used to process data and output scoring results. The results output module is used to output the FDG-PET metabolic load assessment results of the subjects.

10. A method for assessing the metabolic burden of tuberculosis patients using FDG-PET, characterized in that, The method is implemented using the system described in claim 9, and includes the following steps: a) Obtain blood samples from the subjects; b) Detect the expression levels of the blood transcriptome markers in blood samples; c) Input the expression level data obtained in b) into a trained machine learning scoring model, which outputs the subject's FDG-PET metabolic load assessment results. The system as described in claim 9 or the method as described in claim 10, characterized in that, The machine learning scoring model is a linear scoring model constructed based on the gene expression levels of the blood transcriptome markers: PET2 = z(GAS6) + z(GBP6), where z represents the z-score normalization of the expression level of the corresponding gene.

Images